Human stem-cell-derived sensory neurons responded differently to three antibody-drug conjugates in a laboratory assay, with sacituzumab govitecan showing increased electrical activity and the other two showing no consistent change in firing rate. The study paired multi-electrode-array recordings with payload measurements after exposure. At the main late reading, 168 hours, sacituzumab govitecan increased firing at most tested concentrations, while brentuximab vedotin and trastuzumab emtansine did not show significant firing-rate changes. The results describe associations under the test conditions, not proof that one component caused another.
A large gap at the endpoint
The sharpest difference appeared in the amount of payload measured at the endpoint. The assay found SN-38 from sacituzumab govitecan in both the culture medium and cell lysates, with more detected as the concentration rose. It found little or no quantifiable payload from the two comparator ADCs. At the highest tested concentrations, the measured-to-theoretical payload ratio was 94.6% for sacituzumab govitecan and below 0.2% for both brentuximab vedotin and trastuzumab emtansine at 168 hours. This ratio compares the total amount measured with the amount expected in theory, so it should be read as an endpoint measurement rather than a complete record of release over time.
Inside the neuron assay
The platform tracked spontaneous neuronal firing with a multi-electrode array, a plate that records electrical spikes from many points at once. Each recording lasted 15 minutes. An electrode counted as active when it fired at least 5 spikes per minute, and a well would have been excluded if fewer than 9 of its 16 electrodes were active. No wells met that exclusion criterion. The main analyses used six wells per condition, with the same wells followed over time. The exposure protocol used matched vehicle controls, and half the medium was replaced with compound-containing medium after the 72-hour measurement.
Free payloads move in opposite directions
MMAE and Lys-SMCC-DM1 were associated with lower firing frequency after 72 hours, especially at the highest concentration. SN-38 showed the opposite pattern, with higher firing rate and mean spike amplitude at 168 hours at every concentration tested. Higher-concentration payload exposure was also associated with increased resistance. At dose 4, MMAE and Lys-SMCC-DM1 were associated with marked neurite fragmentation and disruption of the neuronal network, while SN-38 showed no obvious morphological change even at its highest tested concentration.
An independent experiment using a different neuronal lot reproduced a fall in spike rate with MMAE and a rise with SN-38. The longitudinal outcomes were analyzed with repeated-measures mixed-effects models that accounted for time, dose, their interaction, plate and well-level effects. Generalized estimating equations were used when mixed models did not converge or became singular, followed by Dunnett-type adjusted comparisons.
A signal without a clear surface target
Target staining added a second caution. In human peripheral nerve sections, qualitative immunostaining found no clear neuronal or Schwann-cell signal for CD30, HER2 or Trop2. The iPSC-derived sensory neurons likewise showed no clear membrane-associated signal for those targets, although limited intracellular or non-neuronal signals were seen. The staining therefore did not provide a clear surface-target explanation for the activity patterns, but a qualitative absence of a clear signal is not the same as proving that no target is present.
A useful screen, with limits
The authors present the combined readout as a complementary tool for early ADC evaluation, bringing together neuronal activity, the assay's resistance measure, morphology and measured payload. It shows that free payloads and corresponding ADCs can produce different profiles under the stated conditions. It does not, however, establish that measured SN-38 caused sacituzumab govitecan's electrical changes, and the increase in firing should not be read as proof of toxicity.
This is a laboratory comparison under defined exposure conditions. The key payload values were measured at the 168-hour endpoint, while the functional comparisons were condition-specific. The findings therefore support an integrated assay for ADC evaluation, but they do not by themselves establish a causal path from released payload to neuronal response.
Study disclosures
The human-tissue component used commercially available, de-identified dorsal-root-ganglion sections, and the authors state that ethics approval and informed consent were not required under applicable institutional policies. The study reports support from JSPS KAKENHI grant JP23K06286 and AMED grant JP25be1004103. Several authors reported employment or founder/employee relationships with Eisai Co., Ltd. or VitroVo Inc.; the remaining authors reported no relevant commercial or financial relationships, and a reviewer disclosed past co-authorship with two authors. The authors state that raw data supporting the conclusions will be made available without undue reservation.
Paper data and sources
Original title: Integrated electrophysiological and payload-release assessment of neuronal responses to antibody-drug conjugates in human iPSC-derived sensory neurons: implications for in vitro neurotoxicity evaluation
Authors: Kuroda T, Han X, Mochizuki T et al.
Journal/Repository: Frontiers in toxicology
Status: Peer-reviewed
First online: 2026-10-09
DOI: Not available
Original paper · Full text